Abstract: To effectively evaluate the lubricated state of a sliding bearing and prevent serious accidents caused by sliding bearing faults, a diagnosis method was proposed for the lubrication state of a sliding bearing based on information entropy and information entropy distance. The dry friction state, boundary friction state and liquid friction state of the sliding bearing were simulated on a 300 MW turbo-generator test rig, and subsequently acoustic emission signals of different friction states were obtained, which were then analyzed to identify the lubrication state using the information entropy distance method, so as to guarantee the operation performance and safety of the sliding bearing. Results show that the information entropy distance diagnosis method with wavelet space feature spectrum entropy is more accurate than that without wavelet space feature spectrum entropy.